US government tightens grip on AI model releases, threatens industry

The U.S. government is tightening its grip on frontier AI models, with both Anthropic and OpenAI facing release delays. Two weeks after the government pulled Anthropic’s Fable and Mythos models, OpenAI’s new GPT 5.6 is headed for a similar limbo. According to The Information, GPT 5.6 will launch only in a limited preview, with the government approving each customer individually until a general release is authorized. Sam Altman reportedly hopes the preview will last only a couple of weeks, but Anthropic’s Mythos has already been in preview for months with no clear path to a general release. Even short delays could significantly reduce the economic returns on costly AI systems, at a time when labs are desperate to improve their bottom lines. If model development slows, it could also chill the ongoing massive data center buildout, putting the entire industry at risk.

The core problem, as GMU fellow Dean Ball explains, is that the U.S. government lacks the expertise and capacity to perform the kind of testing needed for frontier models. Regulators have not articulated what specific risks they are trying to mitigate, nor what safety assurances would satisfy them. While some regulation is reasonable (as with consumer products), the current ad hoc approval process creates uncertainty for all labs. The article argues that the industry must now work together—trusting independent groups to guide the process, supporting the least-bad regulatory options, and fighting for AI as a collective rather than using safety as a competitive weapon. With AI capabilities now having real political consequences, collective action is essential for the industry to survive and thrive.